# US 2024/0137258 A1 · Method and device(s) for supporting machine learning based crest factor reduction and digital predistortion

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2024/0137258 A1
- Document ID: US20240137258A1
- Status: Published application
- Kind: A1
- Application number: 18/277,318
- Filed: 2021-02-22
- Published: 2024-04-25
- Earliest claimed: 2021-02-22
- Assignees: Telefonaktiebolaget LM Ericsson (publ)
- Applicants: Telefonaktiebolaget LM Ericsson (publ)
- Inventors: S M Shahrear TANZIL; Ashim BISWAS
- Web page: https://patentgrep.com/patent/US20240137258A1

## CPC

- H04L 27/2623: {Reduction thereof by clipping}
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/045: Combinations of networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G06N 3/09: Supervised learning
- H03F 1/3247: {using feedback acting on predistortion circuits (H03F1/3264 takes precedence)}
- H03F 1/3252: {using multiple parallel paths between input and output (H03F1/3258, H03F1/3282, H03F1/3294 take precedence)}
- H03F 3/19: with semiconductor devices only
- H03F 3/245: {with semiconductor devices only}
- H04L 27/367: {using predistortion}
- G06N 20/00: Machine learning
- H03F 2200/451: the amplifier being a radio frequency amplifier
- H03F 2201/3206: Multiple channels are combined and amplified by only one amplifier
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H03F: AMPLIFIERS

## Abstract

Method and device(s) for supporting performance of machine learning based CFR and DPD on multiple digital input signals relating to different frequency bands, respectively, in order to signal condition said signals before power amplification and subsequent transmission in said frequency bands by a wireless communication network. The device(s) obtain said multiple digital input signals as complex valued signals. The device(s) perform feature construction that takes said multiple digital input signals as input and provides constructed feature signals according to predefined constructed feature types as output. Said predefined constructed feature types relate to at least the following per complex valued sample of the obtained complex valued multiple digital input signals the real part of the sample, the imaginary part of the sample and at least one of the absolute value of the sample and the phase of the sample.

## Claims

### Claim 1

A method, performed by a device, for supporting performance of machine learning based crest factor reduction, CFR, and/or digital predistortion, DPD, on multiple digital input signals relating to different frequency bands, respectively, in order to signal condition said signals before power amplification and subsequent transmission in said frequency bands by a wireless communication network, the method comprising:
obtaining said multiple digital input signals as complex valued signals; and
performing feature construction that takes said multiple digital input signals as input and provides constructed feature signals according to predefined constructed feature types as output, wherein said predefined constructed feature types relate to at least the following per complex valued sample of the obtained complex valued multiple digital input signals: the real part of the sample, the imaginary part of the sample and at least one of the absolute value of the sample and the phase of the sample.

### Claim 2 (depends on 1)

The method of claim 1, wherein said feature construction takes said multiple digital input signals as input and per present complex valued sample of each signal provides at least the following constructed feature signal samples as constructed feature output associated each present complex valued sample of the signal: the real part of the present sample, the imaginary part of the present sample and at least one of the absolute value of the present sample and the phase of the present sample.

### Claim 3 (depends on 2)

The method of claim 2, wherein the feature construction for each of said present complex valued samples also considers one or more preceding complex valued samples and provides, for each such preceding complex valued sample, at least the following constructed feature signal samples as further output associated with said present complex valued sample: the real part of the preceding sample, the imaginary part of the preceding sample and at least one of the absolute value of the preceding sample and the phase of the preceding sample.

### Claim 4 (depends on 1)

The method of claim 1, wherein the output of constructed feature signals corresponds to samples comprised in one or more 1-dimensional data structures with data grouped per digital input signal of said multiple digital input signal.

### Claim 5 (depends on 1)

The method of claim 1, wherein the output of constructed feature signals corresponds to samples comprised in a 2-dimensional data structure with number of a total of samples considered for each digital input signal in one dimension and the constructed feature signal samples for all the multiple digital input signals in the other dimension.

### Claim 6 (depends on 1)

The method of claim 1, wherein the method further comprises:
applying CFR and/or DPD signal conditioning based on one or more trained machine learning models implementing said machine learning based CFR and/or DPD, wherein each of said trained machine learning models takes said constructed feature signals as input, and said trained machine learning models provide CFR and/or DPD first signal conditioned multiple digital output signals as output.

### Claim 7 (depends on 6)

The method of claim 6, wherein the method further comprises:
applying said power amplification and said wireless transmission based on the signal conditioned multiple digital output signals, wherein said power amplification is performed with a single power amplifier for all of said different frequency bands.

### Claim 8 (depends on 6)

The method of claim 6, wherein the CFR and/or DPD signal conditioning comprises both CFR and DPD.

### Claim 9 (depends on 8)

The method of claim 8, wherein the CFR and DPD are separately and subsequently applied.

### Claim 10 (depends on 9)

The method of claim 9, wherein the CFR operates on said received multiple digital input signals and provides CFR multiple digital output signals, wherein the DPD subsequently operates on said CFR multiple digital output signals and provides said signal conditioned multiple digital output signals.

### Claim 11 (depends on 10)

The method of claim 10, wherein both the CFR and DPD are based on machine learning.

### Claim 12 (depends on 8)

The method of claim 8, wherein the CFR and DPD are jointly applied with said one or more trained machine learning models being common for both the CFR and the DPD, wherein the jointly applied CFR and DPD operates on said received multiple digital input signals and provides said first signal conditioned multiple digital output signals.

### Claim 13 (depends on 6)

The method of claim 6, wherein said CFR and/or DPD signal conditioning comprises said machine learning based DPD and said method further comprises:
applying further DPD signal conditioning comprising an additional, tracking, DPD that operates on the error from said machine learning based DPD and provides DPD second signal conditioned multiple digital output signals as output.

### Claim 14 (depends on 12)

The method of claims 12, wherein said one or more trained machine learning models that are common for both the CFR and the DPD additionally provide CFR signal conditioned multiple digital output signals to be used as input to said tracking DPD for the provision of said residual error.

### Claim at position 15

15-16. (canceled)

### Claim 17

A device for supporting performance of machine learning based crest factor reduction, CFR, and/or digital predistortion, DPD, on multiple digital input signals relating to different frequency bands, respectively, in order to signal condition said signals before power amplification and subsequent transmission in said frequency bands by a wireless communication network, the device configured to:
obtain said multiple digital input signals as complex valued signals; and
perform feature construction that takes said multiple digital input signals as input and provides constructed feature signals according to predefined constructed feature types as output, wherein said predefined constructed feature types relate to at least the following per complex valued sample of the obtained complex valued multiple digital input signals: the real part of the sample, the imaginary part of the sample and at least one of the absolute value of the sample and the phase of the sample.

### Claim 18 (depends on 17)

The device of claim 17, wherein said feature construction takes said multiple digital input signals as input and per present complex valued sample of each signal to provide at least the following constructed feature signal samples as constructed feature output associated each present complex valued sample of the signal: the real part of the present sample, the imaginary part of the present sample and at least one of the absolute value of the present sample and the phase of the present sample.

### Claim 19 (depends on 18)

The device of claim 18, wherein the feature construction for each of said present complex valued samples also considers one or more preceding complex valued samples and to provide, for each such preceding complex valued sample, at least the following constructed feature signal samples as further output associated with said present complex valued sample: the real part of the preceding sample, the imaginary part of the preceding sample and at least one of the absolute value of the preceding sample and the phase of the preceding sample.

### Claim 20 (depends on 17)

The device of claim 17, wherein the output of constructed feature signals corresponds to samples comprised in one or more 1-dimensional data structures with data grouped per digital input signal of said multiple digital input signal.

### Claim 21 (depends on 17)

The device of claim 17, wherein the output of constructed feature signals corresponds to samples comprised in a 2-dimensional data structure with number of a total of samples considered for each digital input signal in one dimension and the constructed feature signal samples for all the multiple digital input signals in the other dimension.

### Claim 22 (depends on 17)

The device of claim 17, wherein the device is further configured to:
apply CFR and/or DPD signal conditioning based on one or more trained machine learning models implementing said machine learning based CFR and/or DPD, wherein each of said trained machine learning models takes said constructed feature signals as input, and said trained machine learning models provide CFR and/or DPD first signal conditioned multiple digital output signals as output.

### Claim at position 22

23-30. (canceled)
